Skip to content

What do you use for configuration drift detection?

General Discussion by tomhider 22 replies 2.5K views
#11

I pay someone in Nairobi to check my shop's server weekly. Cheaper than my time and he WhatsApps me a screenshot of what changed. Low tech but my uptime is better than when I used fancy tools I didn't understand.

#12

tomhider said:
My inbox died
Use a webhook to Matrix or Slack instead of email. I have a 5-line bash script on my Osaka Vultr box that diffs /etc against git and posts to a private Matrix room if anything changed. Took 20 minutes to set up.

#13

In France we have this saying: the best tool is the one you will actually use. For my Bordeaux VPS I just set up a simple cron with aide, it emails once a week with a summary. Not instant but I don't need instant for a personal site.

#14

FyreJun said:
Posts to a private Matrix room
This is actually smart. I use matrix anyway. Is this just like git diff piped to curl or somethign more fancy?

#15

tomhider said:
Just like git diff piped to curl
Basically yes. I use git diff --stat to keep it short, and a simple if [ -n "$output" ]; then curl with the text. The full diff I can run manually if the stat looks weird.

#16

I tried OSQuery for this on my home server. Massive overkill. Felt like bringing a Kubernetes to a knife fight. Uninstalled after a week.

#17

ben20 said:
Etckeeper
Seconding this. Found it on a scrapped Dell server I bought from a Latvian university auction, already configured. Stole the idea for my own boxes. Zero maintenance.

worst bandwidth, best stories
#18

ben20 said:
Etckeeper
FyreJun said:
I use git diff --stat to keep it short, and a simple if ; then curl with the text.
Okay I think I understand now. I don't need a "configuration drift detection platform," I just need to not forget what I changed. Going to try etckeeper first since it's in the Vultr Ubuntu image already.

learning on $5 VPS and prayers
#19

sarahjey said:
I just need to not forget what I changed
This is the real insight. We all pretended we needed enterprise solutions for our toy servers. My notes.txt was actually fine, I just needed to back it up somewhere.

RAM is just fast disk, right?
#20

For my inference boxes in San Jose I use a slightly cursed solution: I snapshot the NVMe with dd to a cheap USB drive before any big change. If I drift too far I just dd back. 15 minutes to restore, zero config management overhead. Not recommended for production.

CUDA cores are my love language

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft